Anthropic warns of AI risks / Models could be used for biological weapons and cyberattacks

Anthropic warns in a recent threat report that frontier artificial intelligence models can be used in activities supporting the development of biological weapons, as well as in cyberattacks, influence operations, surveillance, scams and fraud. The company says it blocked researchers who used AI in ways that could have contributed to the development of biological weapons.

The report presents five cases in which Anthropic models were allegedly used for activities with abuse potential. The company considers malicious biological use one of the most serious risks associated with advanced artificial intelligence systems and warns that, without adequate safety measures, the consequences could be catastrophic.

Over the past eight months, researchers have identified cases ranging from fake dating apps and scams through hotel Wi-Fi networks to surveillance systems designed to identify dissidents. Anthropic’s Claude model was also allegedly used in a cyberespionage campaign linked to Russia and an Iranian propaganda institution.

Similar warnings have been issued by other companies. Google announced that an individual had attempted to use Gemini to obtain a technical guide on synthesizing biological agents weaponized for use as weapons. U.S. Senator Bernie Sanders called for a pause in the development of advanced AI and a ban on artificial superintelligence.

President Donald Trump has so far dismissed these concerns, arguing that the United States would be in a highly vulnerable position if it failed to win the global competition in artificial intelligence.